- Email: [email protected]
The influence of weather on human mortality in Hong Kong
Social Science & Medicine 50 (2000) 419±427
www.elsevier.com/locate/socscimed
The in¯uence of weather on human mortality in Hong Kong Yuk Yee Yan* Department of Geography, Hong Kong Baptist University, Kowloon Tong, Hong Kong, People's Republic of China
Abstract This study is the ®rst attempt to investigate mortality seasonality and weather±mortality relationships in Hong Kong from 1980 to 1994. Monthly mortality data from all causes of death, neoplasm, circulatory and respiratory diseases were obtained from the Census and Statistics Department and the weather data were obtained from the Hong Kong Observatory. Regression analyses and ANOVA were employed. Signi®cant winter peaks in sex speci®c and total deaths from all causes, circulatory and respiratory diseases were ascertained. Cancer mortality, however, was not seasonal. Mortality seasonality only existed in age groups 45±64 and r65. For the impact of weather on mortality, no signi®cant relationship between weather variables and cancer mortality was observed. A signi®cant negative association between minimum temperature and a positive relationship between cloud and deaths were found. This suggests that colder and cloudy conditions may heighten mortality. Wind was discovered to have a negative association with mortality. This ®nding revealed that the stressful eect of wind on mortality was negligible. There was no apparent sex dierence. Deaths from the younger age groups (0±24 yr old) were not weather related. Weak weather connection with mortality for age group 25±44 was discovered, with Adj r 2 values ranging from 0.05 to 0.07. The elderly (age r65) were more vulnerable to weather stress and strong weather± mortality relationship was uncovered, with Adj r 2 values from 0.36 to 0.66. These results are important information for formulating public health policies. # 1999 Elsevier Science Ltd. All rights reserved. Keywords: Seasonality; Mortality; Hong Kong
Introduction The awareness of the relationship between climate and human health has long been existed since Hippocrates. This association has been examined by researchers from various disciplines (Mather, 1974; Oliver, 1981). There are growing interests in the investigation of seasonal variations of mortality and weather±mortality relationship by the medical and climatological disciplines.
* Tel.: +852-2339-7166; fax: +852-2339-5990. E-mail address: [email protected] (Y.Y. Yan)
It is known that mortality is generally higher in winter (Tromp, 1963; Hodge, 1978; Bako et al., 1988; Douglas et al., 1991a, 1991b; Lerchl, 1998; McKee et al., 1998). Kalkstein and Davis (1989) found that warm, humid and calm conditions were related to elevated summer mortality whereas cloudy, damp and snowy conditions were associated with the greatest mortality in winter. Shumway et al. (1988) showed that temperature, but not relative humidity, contributed signi®cantly to mortality. Alberdi et al. (1998) discovered a J-shaped relationship between temperature, relative humidity and mortality in Madrid; and mortality was found to be negatively associated with high humidity in summer.
0277-9536/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 9 5 3 6 ( 9 9 ) 0 0 3 0 1 - 9
420
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
The impact of stressful weather on mortality is well documented (e.g. Marmor, 1975; Bridger et al., 1976; Ellis and Nelson, 1978; Applegate et al., 1981; Ramlow and Kuller, 1990; Knobeloch et al., 1997; Whitman et al., 1997; Smoyer, 1998). Gover (1938) stated that excessive mortality during a second heat wave would be slight in comparison to excessive mortality during the ®rst. Buechley et al. (1972) discovered that maximum temperature over 358C (958F) was associated with excess mortality. Ellis (1972) suggested that excessive heat might be more lethal than elevated air pollution. Saez et al. (1995) claimed that mortality would increase following at least three consecutive days of increased temperatures. Research has also been conducted on the impact of extreme winter weather on mortality (e.g. Gerald and Rose, 1979; Kunst et al., 1993; Christophersen, 1997; Donaldson and Keatinge, 1997). It has been documented that the eect of low temperature was more apparent in population in warm regions and more deaths followed uncommon cold weather (Larson, 1990; Makino, 1993). Little research on climate and mortality has been done in the tropics. Madrigal (1994) reported maximum mortality during the rainy season in Costa Rica Motohashi et al. (1996) investigated the relationship between rainfall and seasonal mortality in Sri Lanka.
Table 1 Weather variables used in the study Maximum Temperature (MAXT) Minimum Temperature (MINT) Dewpoint (DEWP) Pressure (PRES) Cloud (CLD) Windspeed (WINDSP) Precipitation (PRECIP)
Auliciems et al. (1997) detected an increase in mortality with cold, dry weather in Brisbane, Australia. Kalkstein and Smoyer (1993) found that the greatest mortality in Guangzhou was signi®cantly associated with hot afternoon temperatures and very warm night temperatures. Pan et al. (1995) studied cardiovascular disease mortality in Taiwan and concluded that there was a U-shaped relation between temperature and mortality from coronary artery disease and cerebral infarction. Research by Woo et al. (1991) revealed that temperature was not strongly associated with stroke occurrence in Hong Kong. Knowledge about the interaction between weather and mortality can provide better understanding about the vulnerability of speci®c population groups. The present study provides an overview of weather±mor-
Fig. 1. Various causes of death.
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
421
Fig. 2. Average monthly maximum, minimum temperatures and rainfall 1980±1994.
tality association in Hong Kong. The objectives of this research are to test the hypothesis that there was seasonality of mortality for the period from 1980 through 1994 and to assess weather±mortality relationships for genders and various age groups.
Data and method Data Monthly mortality from various diseases in this study period were obtained from the Census and Statistics Department. The data were further divided into gender and age groups 0±16, 17±24, 25±44, 45±64 and r65. Current knowledge suggests that there are associations between various weather variables and mor-
tality. Weather variables including temperature, dewpoint, pressure, cloud, windspeed and precipitation were selected for this analysis and monthly meteorological data for the study period were obtained from the Hong Kong Observatory (Table 1). Figure 1 presents the average monthly maximum and minimum temperatures and rainfall for the study period. Temperatures were close to the normal. However, the summer rainfall was below normal. Rainfall anomalies were observed in the mean May and August precipitation ®gures. The May average of 356.4 mm exceeded the norm of 316.7 mm and the August values was 328.3 mm as compared with the norm of 391.4 mm. Determination of mortality seasonality There were a total of 410,281 deaths in the study area and the causes of death from neoplasm, circulatory and respiratory systems failure comprised 29.2, Table 2 Selected causes of death
Fig. 3. Monthly mortality of sex speci®c and total deaths from all causes.
Cause of death
ICD section
ICD codes
All causes Neoplasm (malignant) Circulatory system Hypertensive disease Ischaemic heart disease (IHD) Cerebrovascular disease (CVD) Other circulatory diseases Respiratory diseases
I±XVII II VII
inclusive 140±208 390±459 401±405 410±414 430±438
VIII
460±519
422
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
Table 3 Actual numbers of deaths of selected diseases of the circulatory system
Male Female Age 25±44 Age 45±64 Age r65
Hypertensive disease
IHD
CVD
Other circulatory
6672 6155 214 3319 9289
21,596 18,344 666 9077 30,163
22,394 24,674 1141 11674 33,979
9099 8981 1237 4733 11,936
28.8 and 16.8% respectively (Fig. 2). Only the major causes of death were selected for analysis (Table 2). The greatest numbers of deaths from circulatory system were the hypertensive disease, ischaemic heart disease (IHD) and cerebrovascular disease (CVD) (Table 3). Monthly gender and age speci®c and total mortality data were divided by their corresponding annual midyear population to eliminate the eect of population growth. Further, these adjusted monthly mortality data were then standardized to a 30.5 day average to correct for monthly dierences in number of days. Winter peaks existed in total and gender and age speci®c mortality from all causes (Figs. 3±5). In order to test the existence of mortality seasonality, dummy variables (winter months, i.e. January, February and December=1, while other months=0) were employed in regression analyses. ANOVA was then utilized to test for dierences in seasonal mortality. Determination of weather±mortality relationship Regression analysis was used to ascertain which combination of weather elements would produce the highest coecient determination (r 2) for age and gender speci®c deaths and total mortality from all causes and selected diseases. Each meteorological variable was
Fig. 4. Monthly mortality of sex speci®c deaths from selected causes.
Fig. 5. Monthly mortality of age r65 from selected causes.
checked by inspection of scatterplots for a nonlinear relation to mortality. When non-linearity was evident, appropriate transformation was employed. In this analysis, maximum temperature, minimum temperature and dewpoint were logarithmically transformed, whereas a square root transformation was chosen for precipitation. Then a stepwise approach was adopted. Regression diagnostics including residual plots and variance in¯ation factors (VIFs) (Draper and Smith, 1981) were used to choose the ®nal model. A high VIF indicates the inclusion of collinear independent variables in the model. When this occurs, one of the collinear variables is omitted. The ®nal selected model must have VIF for all independent variables less than 2.00 and randomly distributed residuals.
Results Mortality seasonality Results of the regression analyses using dummy variables (Table 4) revealed a positive relationship between mortality and the winter months, i.e. higher deaths occurred in winter. Statistically signi®cant reliable winter peaks in gender speci®c and total deaths from all causes, circulatory and respiratory diseases were noted. However, cancer mortality was not seasonal. There was no gender dierence in mortality seasonality. Excepting neoplasms, results from ANOVA (Table 5) exhibited signi®cant dierences of mortality among seasons and this further con®rmed the existence of a reliable winter peak for all causes and selected disease mortality. For the various age groups, no signi®cant seasonality for neoplasm was discovered. Only age groups 45± 64 and r65 displayed signi®cant winter peaks in deaths from circulatory and respiratory systems. Signi®cant seasonality for all causes of deaths existed only in age group r65. The winter and summer mortality rates for
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
423
Table 4 Models for mortality seasonality (winter peak) Total death
All causes Neoplasm Circulatory Hypertensive IHD CVD Other circulatory Respiratory
Male
Female
Age r65
regression coecient
p
regression coecient
p
regression coecient
p
regression coecient
p
6.34 0.26
0.00 0.14
6.44 0.33
0.00 0.17
6.23 0.19
0.00 0.57
73.71 3.03
0.00 0.07
0.50 1.06 1.04 0.51 1.76
0.00 0.00 0.00 0.00 0.00
0.46 1.09 1.04 0.48 2.05
0.00 0.00 0.00 0.00 0.00
0.55 1.05 1.01 0.56 1.46
0.00 0.00 0.00 0.00 0.00
5.11 11.56 10.78 6.22 18.76
0.00 0.00 0.00 0.00 0.00
the age speci®c deaths, presented in Table 6, further reveals the minimal seasonal mortality dierences among the younger age groups. Further, for the age groups below 44, the summer mortality rates from all causes were found to be slightly higher than the winter ones. The same fact happened in age group 25±44 for IHD, CVD and other circulatory causes. Weather±mortality relationship The results of the regression analyses were shown in Table 7. No signi®cant relationship was detected between weather variables and cancer mortality for total, gender and age speci®c deaths. Minimum temperature had the most signi®cant impact on deaths from the selected causes. For mortality from all causes, a signi®cant negative relationship between minimum temperature and a positive association between cloud cover and gender speci®c and total deaths were observed. However, the contributing weather variables aecting female mortality from all causes were slightly dierent from those of male mortality. Minimum temperature, cloud cover
and windspeed were included in the female model while the male model only contained minimum temperature. Minimum temperature and cloud had the most in¯uential eects on deaths from the circulatory system. Windspeed was discovered to have a signi®cant negative association with deaths from CVD. Deaths from respiratory diseases were also connected with cold and cloudy conditions. There was no prominent sex dierence in mortality from these causes. The evaluation of weather impact on mortality for various age groups revealed that no relationship was observed for the two youngest age groups (ranging from 0 to 24 yr old). For age group 25±44, deaths from IHD and other circulatory diseases were not weather related. The negative eect of windspeed was found to be more notable than the eect of minimum temperature as it was included in all models. The weather±mortality relationships were discovered to be weak (Adj r 2 values ranged from 0.05 to 0.07). Deaths for age groups 45±64 and r65 were mainly contributed by the combined force of minimum temperature and cloud cover. The negative in¯uence of windspeed was
Table 5 Results of ANOVA Total death
All causes Neoplasm Circulatory Hypertensive IHD CVD Other circulatory Respiratory
Male
Female
Age r65
F Ratio
p
F Ratio
p
F Ratio
p
F Ratio
p
63.86 0.91
0.00 0.35
47.87 1.19
0.00 0.32
66.55 0.67
0.00 0.57
45.81 1.88
0.00 0.13
31.16 40.26 49.23 31.17 46.78
0.00 0.00 0.00 0.00 0.00
22.65 32.14 41.05 26.15 37.51
0.00 0.00 0.00 0.00 0.00
31.31 34.15 39.64 27.58 48.12
0.00 0.00 0.00 0.00 0.00
31.57 62.24 33.61 23.76 55.63
0.00 0.00 0.00 0.00 0.00
424
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
Table 6 Mortality rates (per 100,000) for age speci®c groups All causes Age 0±16 Winter Summer Age 17±24 Winter Summer Age 25±44 Winter Summer Age 45±64 Winter Summer Age r65 Winter Summer
Hypertensive disease
IHD
CVD
Other circulatory
Respiratory
262.8 288.5
0.0 0.0
0.0 0.0
2.3 2.3
5.9 5.9
26.6 24.3
170.6 180.0
0.0 0.0
1.0 1.0
5.0 3.6
7.7 6.3
13.1 11.7
384.3 393.3
4.1 3.2
8.9 9.5
15.3 16.2
17.1 17.6
26.1 23.0
3059.6 2858.0
100.4 72.9
257.4 202.5
333.5 265.5
72.0 60.8
365.9 302.9
16356.6 12572.6
707.9 429.8
2108.7 1494.9
2338.7 1757.7
898.7 572.0
3641.4 2724.8
Fig. 6. Scatterplot of cerebrovascular (CVD) mortality (total death) versus minimum temperature.
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
425
Table 7 Models for selected causes of deatha Explanatory variables Total death MINT CLD WINDSP Adj r 2 Male MINT CLD WINDSP Adj r 2 Female MINT CLD WINDSP Adj r 2 Age 25±44 MINT CLD WINDSP Adj r 2 Age 45±64 MINT CLD WINDSP Adj r 2 Age r65 MINT CLD WINDSP Adj r 2 a b
All causes ÿ33.06 4.35 0.61 ÿ33.83 0.53 ÿ34.03 7.51 ÿ0.92b 0.62
Hypertensive disease ÿ2.64 0.36b 0.44 ÿ2.53 0.38b 0.35 ÿ2.83 0.49b 0.45
IHD ÿ5.36 0.48 ÿ5.49 0.41 ÿ5.27 0.44
CVD
Other circulatory
ÿ5.65 1.40 ÿ0.24 0.59
ÿ2.87 0.53
ÿ9.32 2.24
0.43
0.46
ÿ5.59 0.98b ÿ0.33 0.58
ÿ2.57 0.40b
ÿ5.61 1.84 ÿ0.30b 0.44
ÿ3.24 0.68b
0.33
0.41
ÿ0.12b ÿ0.70b 0.05
ÿ0.02b 0.05
ÿ27.07b
ÿ2.74 0.93b
0.04 ÿ378.14 45.04 0.49
0.17 ÿ28.00 4.84b ÿ1.14b 0.47
0.29 ÿ59.28 0.66
ÿ6.93 2.16b ÿ1.06 0.21 ÿ60.13 17.13 ÿ3.87b 0.48
ÿ10.85 2.47 0.44 ÿ8.09 2.67 0.44 ÿ0.34b 0.28b ÿ0.10b 0.06
ÿ0.08b 0.07 ÿ4.84
Respiratory
ÿ1.61b 0.92b
ÿ8.39 2.83b
0.06
0.24
ÿ32.32
ÿ101.47 24.24
ÿ1.77b 0.36
0.54
Numbers represent regression coecient (b) at p = 0.00. Numbers represent regression coecient (b) at p < 0.05.
found to be more prominent for age group r65. Much stronger weather±mortality associations were also detected in the elderly (Adj r 2 values ranged from 0.36 to 0.66).
Discussion The ®ndings that mortality seasonality existed for gender speci®c and total deaths are consistent with previous studies (Bako et al., 1988; Douglas et al., 1991a, 1991b). Although Allan (1966) found that the highest mortality from cancer in Britain was not in the coldest months, but in the fourth quarter of the year, the result of this study indicated no seasonality for malignant disease. This ®nding con®rms the absence of a winter peak. The non-existence of signi®cant seasonality for the
younger age groups (0±44) was further supported by the ®ndings of this study that deaths of those groups (0±24) were not weather related. Further, this also evidences that the younger age groups are less vulnerable to weather eects. For the weather±mortality relationships, it is not surprised that maximum temperature was excluded from all models. Majority of the population have acclimatized the hot, sultry summer in the subtropics. Any temperature changes, particularly cold weather change, will have more obvious eect and trigger more deaths (Makino, 1993; Auliciems et al., 1997). The negative relationship and positive connection between minimum temperature and cloud cover respectively and mortality from all causes suggest that overcast and colder days may provide conditions for increased mortality. Windspeed was not included in all the models. This shows that the stressful eect of wind
426
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
in colder weather has less in¯uence on mortality as expected and con®rms the previous ®ndings by Kalkstein and Davis (1989). It is not unexpected to discover no prominent gender dierence in mortality from the various causes. However, the inclusion of windspeed and cloud cover in the female model from all causes of death indicated that females were more sensitive to the combined eect of weather stress. Minimum temperature and cloud were the contributing factors to deaths from circulatory system failures. The ®nding of a negative association between windspeed and CVD mortality further reveals the eect of wind on mortality is negligible. Minimum temperature was the most in¯uential factor. It is found that exposure to the cold increases blood pressure, blood viscosity and heart rate (Kunst et al., 1993) and may precipitate circulatory diseases. Pan et al. (1995) stated that there was a U-shaped relation between temperature and cardiovascular mortality in Taiwan. However, no such association was observed in Hong Kong (Fig. 6). Deaths from respiratory diseases were connected with minimum temperature and cloud cover. The eect of colder temperature on the respiratory system has an important role. The alveoli are embedded in a capillary network. The alveolar cells and the capillary endothelial cells comprise two layers of cells between the air and blood. A continuous sheet of blood is close to the alveolar wall. The lungs are exposed directly to the environment without the protection of skin. This intimate environmental contact induces not only respiratory diseases but also provokes changes in circulatory and other diseases (Douglas et al., 1991b). The strong weather±mortality associations in age group r65 and no or very weak relationships for the three youngest age groups (ranging from 0 to 44 yr old) are important discoveries. They are consistent with previous research ®ndings that the elderly are more susceptible to weather stress. This apparent weather±mortality relationship is due to the failure of homeostatic defence mechanism with advancing age that facilitates the onset of hyperthermia or hypothermia, which in turn would trigger circulatory and other diseases (Douglas et al., 1991b; Pan et al., 1995). Further, Woo et al. (1991) also found that the elderly might be more vulnerable to less extreme temperature changes in subtropical climate. Conclusions It is apparent that weather has a substantial eect on mortality. The impact can dier seasonally. Also certain age groups seem to be more susceptible to the in¯uence of weather. Results of this study con®rm some previous ®ndings, but also contradict other ®nd-
ings in weather±mortality research. More attention has recently been placed on the impact of air pollution on mortality and many studies suggest that pollution has a signi®cant eect on health (Stern, 1977). More promising results would be expected if air pollution variables are included. The results also indicate that the situation in Hong Kong is not the same as other tropical or subtropical regions. Rainfall was not associated with mortality as in Costa Rica or Sri Lanka (Madrigal, 1994; Motohashi et al., 1996). Humidity was found to be unrelated to mortality unlike Brisbane (Auliciems et al., 1997). Further research in these regions should be conducted to clarify the complexity of weather±mortality relationships. These research ®ndings contribute to the formulation of public health policy, such as justifying increasing resources for elderly care. Knowledge of weather±mortality relationships may provide better insights on which diseases are more likely to cause death under certain weather conditions and the susceptibility of speci®c age groups. This type of research can aid in the development of health-related weather services to the general public, such as indices to warn of severe temperature or pollution episodes. References Alberdi, J.C., Diaz, J., Montero, J.C., Miron, I., 1998. Daily Mortality in madrid community 1986±1992: relationship with meteorological variables. European Journal of Epidemiology 14, 571±578. Allan, T.M., 1966. Seasonal Distribution of Deaths from Cancer. British Medical Journal 10, 837±841. Applegate, W.B., Runyan Jr., J.W., Bras®eld, L., Willians, M.L., Konigsberg, C., Fouche, C., 1981. Analysis of 1980 the heat wave in Memphis. Journal of American Geriatrics Society 29, 337±342. Auliciems, A., Frost, D., Siskind, V., 1997. The time factor in mortality: weather association in a subtropical environment. International Journal of Biometeorology 40, 183± 191. Bako, G., Ferenczi, L., Hill, G.B., Lindsay, J., 1988. Seasonality of mortality from various diseases in Canada 1979±83. Canadian Journal of Public Health 7, 388±389. Bridger, C.A., Ellis, F.P., Taylor, H.L., 1976. Mortality in St Louis during heat waves in 1936, 1953, 1954, 1955 and 1966. Environmental Research 12, 38±48. Buechley, R.W., Van Bruggen, J., Truppi, L.E., 1972. Heat Island=Death Island? Environmental Research 5, 85±92. Christophersen, O., 1997. Mortality during the 1996/7 winter. Population Trend 90, 11±17. Donaldson, G.C., Keatinge, W.R., 1997. Mortality related to cold weather in elderly people in Southeast England 1979± 94. British Medical Journal 315, 1055±1056. Douglas, A.S., Al-Sayer, H., Rawles, J.M., Allan, T.M., 1991a. Seasonality of disease in Kuwait. Lancet 337, 1393± 1397.
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427 Douglas, A.S., Allan, T.M., Rawles, J.M., 1991b. Composition of seasonality of diseases. Scottish Medical Journal 36, 76±82. Draper, N., Smith, H., 1981. Applied Regression Analysis. John Wiley, New York. Ellis, F.P., 1972. Mortality from heat illness and heat-aggravated illness in the US. Environmental Research 5, 1±58. Ellis, F.P., Nelson, F., 1978. Mortality in the elderly in a heat wave in New York City, August 1975. Environmental Research 15, 504±512. Gerald, F., Rose, R., 1979. Blizzard morbidity and mortality: Rhode Island, 1978. American Journal of Public Health 69, 1050±1052. Gover, M., 1938. Mortality during periods of excessive Temperature. US Public Health Report 53, 1122. Hodge, W., 1978. Weather and mortality. In: EDIS. US Department of Commerce, Washington. Kalkstein, L.S., Davis, R.E., 1989. Weather and human mortality: an evaluation of demographic and interregional responses in the US. Annals of the Association of American Geographers 79, 44±64. Kalkstein, L.S., Smoyer, K.E., 1993. The impact of climate change: some international implications. Experientia 49, 969±979. Knobeloch, L., Anderson, H., Morgan, J., Nashold, R., 1997. Heat-related illness and death, Wisconsin, 1995. Wisconsin Medical Journal 96, 33±38. Kunst, A.E., Looman, C.W.N., MacKenbach, J.P., 1993. Outdoor air temperature and mortality in the netherlands: a time-series analysis. American Journal of Epidemiology 137, 331±341. Larson, U., 1990. Short-term ¯uctuation in death by cause, temperature and income in the United States, 1930±1985. Social Biology 37, 172±187. Lerchl, A., 1998. Changes in the seasonality of mortality in Germany from 1946 to 1995: the role of temperature. International Journal of Biometeorology 42, 84±88. Madrigal, L., 1994. Mortality seasonality in EscazuÂ, Costa Rica 1851±1921. Human Biology 66, 433±452. Makino, K., 1993. Weather/season and death. Asian Medical Journal 36, 580±587.
427
Marmor, M., 1975. Heat wave mortality in New York City, 1949 to 1970. Archive of Environmental Health 30, 130±136. Mather, J.R., 1974. Climatology: Fundamental and Application. McGraw-Hill, New York. McKee, M., Sanderson, C., Chenet, L., Vassin, S., Shkolnikov, V., 1998. Seasonal variation in mortality in Moscow. Journal of Public Health and Medicine 20, 268± 274. Motohashi, Y., Takano, T., Nakamura, K., Nakate, K., Tanaka, M., 1996. Seasonality of Mortality in Sri Lanka: biometeorological considerations. International Journal of Biometeorology 39, 121±126. Oliver, J.E., 1981. Climatology: Selected Applications. John Wiley, New York. Pan, W.H., Li, L.A., Tsai, M.J., 1995. Temperature extremes and mortality from coronary heart disease and cerebral infarction in elderly Chinese. Lancet 345, 353±355. Ramlow, J.M., Kuller, L.H., 1990. Eects of the summer heat wave of 1988 on daily mortality in Allegheny County, PA. Public Health Reports 105, 283±289. Saez, M., Sunyer, J., Castellsague, J., Murillo, C., AntoÂ, J.M., 1995. Relationship between weather temperature and mortality: a time series analysis approach in Barcelona. International Journal of Epidemiology 24, 576±582. Shumway, R.H., Azari, A.S., Pawitan, Y., 1988. Modeling mortality ¯uctuation in los angeles as function of pollution and weather eects. Environmental Research 45, 224±241. Smoyer, K.E., 1998. A comparative analysis of heat waves and associated mortality in St. Louis, Missouri 1980 and 1995. International of Biometeorology 42, 44±50. Stern, A.C., 1977. The Eects of Air Pollution. Academic Press, New York. Tromp, S.W., 1963. Medical Biometeorology. Elsevier, New York. Whitman, S., Good, G., Donoghue, E.R., Benbow, N., Shou, W., Mou, S., 1997. Mortality in Chicago attributed to the July 1995 heat wave. American Journal of Public Health 87, 1515±1518. Woo, J., Kay, R., Nicholls, M.G., 1991. Environmental temperature and stroke in a subtropical climate. Neuroepidemiology 10, 260±265.
www.elsevier.com/locate/socscimed
The in¯uence of weather on human mortality in Hong Kong Yuk Yee Yan* Department of Geography, Hong Kong Baptist University, Kowloon Tong, Hong Kong, People's Republic of China
Abstract This study is the ®rst attempt to investigate mortality seasonality and weather±mortality relationships in Hong Kong from 1980 to 1994. Monthly mortality data from all causes of death, neoplasm, circulatory and respiratory diseases were obtained from the Census and Statistics Department and the weather data were obtained from the Hong Kong Observatory. Regression analyses and ANOVA were employed. Signi®cant winter peaks in sex speci®c and total deaths from all causes, circulatory and respiratory diseases were ascertained. Cancer mortality, however, was not seasonal. Mortality seasonality only existed in age groups 45±64 and r65. For the impact of weather on mortality, no signi®cant relationship between weather variables and cancer mortality was observed. A signi®cant negative association between minimum temperature and a positive relationship between cloud and deaths were found. This suggests that colder and cloudy conditions may heighten mortality. Wind was discovered to have a negative association with mortality. This ®nding revealed that the stressful eect of wind on mortality was negligible. There was no apparent sex dierence. Deaths from the younger age groups (0±24 yr old) were not weather related. Weak weather connection with mortality for age group 25±44 was discovered, with Adj r 2 values ranging from 0.05 to 0.07. The elderly (age r65) were more vulnerable to weather stress and strong weather± mortality relationship was uncovered, with Adj r 2 values from 0.36 to 0.66. These results are important information for formulating public health policies. # 1999 Elsevier Science Ltd. All rights reserved. Keywords: Seasonality; Mortality; Hong Kong
Introduction The awareness of the relationship between climate and human health has long been existed since Hippocrates. This association has been examined by researchers from various disciplines (Mather, 1974; Oliver, 1981). There are growing interests in the investigation of seasonal variations of mortality and weather±mortality relationship by the medical and climatological disciplines.
* Tel.: +852-2339-7166; fax: +852-2339-5990. E-mail address: [email protected] (Y.Y. Yan)
It is known that mortality is generally higher in winter (Tromp, 1963; Hodge, 1978; Bako et al., 1988; Douglas et al., 1991a, 1991b; Lerchl, 1998; McKee et al., 1998). Kalkstein and Davis (1989) found that warm, humid and calm conditions were related to elevated summer mortality whereas cloudy, damp and snowy conditions were associated with the greatest mortality in winter. Shumway et al. (1988) showed that temperature, but not relative humidity, contributed signi®cantly to mortality. Alberdi et al. (1998) discovered a J-shaped relationship between temperature, relative humidity and mortality in Madrid; and mortality was found to be negatively associated with high humidity in summer.
0277-9536/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 9 5 3 6 ( 9 9 ) 0 0 3 0 1 - 9
420
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
The impact of stressful weather on mortality is well documented (e.g. Marmor, 1975; Bridger et al., 1976; Ellis and Nelson, 1978; Applegate et al., 1981; Ramlow and Kuller, 1990; Knobeloch et al., 1997; Whitman et al., 1997; Smoyer, 1998). Gover (1938) stated that excessive mortality during a second heat wave would be slight in comparison to excessive mortality during the ®rst. Buechley et al. (1972) discovered that maximum temperature over 358C (958F) was associated with excess mortality. Ellis (1972) suggested that excessive heat might be more lethal than elevated air pollution. Saez et al. (1995) claimed that mortality would increase following at least three consecutive days of increased temperatures. Research has also been conducted on the impact of extreme winter weather on mortality (e.g. Gerald and Rose, 1979; Kunst et al., 1993; Christophersen, 1997; Donaldson and Keatinge, 1997). It has been documented that the eect of low temperature was more apparent in population in warm regions and more deaths followed uncommon cold weather (Larson, 1990; Makino, 1993). Little research on climate and mortality has been done in the tropics. Madrigal (1994) reported maximum mortality during the rainy season in Costa Rica Motohashi et al. (1996) investigated the relationship between rainfall and seasonal mortality in Sri Lanka.
Table 1 Weather variables used in the study Maximum Temperature (MAXT) Minimum Temperature (MINT) Dewpoint (DEWP) Pressure (PRES) Cloud (CLD) Windspeed (WINDSP) Precipitation (PRECIP)
Auliciems et al. (1997) detected an increase in mortality with cold, dry weather in Brisbane, Australia. Kalkstein and Smoyer (1993) found that the greatest mortality in Guangzhou was signi®cantly associated with hot afternoon temperatures and very warm night temperatures. Pan et al. (1995) studied cardiovascular disease mortality in Taiwan and concluded that there was a U-shaped relation between temperature and mortality from coronary artery disease and cerebral infarction. Research by Woo et al. (1991) revealed that temperature was not strongly associated with stroke occurrence in Hong Kong. Knowledge about the interaction between weather and mortality can provide better understanding about the vulnerability of speci®c population groups. The present study provides an overview of weather±mor-
Fig. 1. Various causes of death.
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
421
Fig. 2. Average monthly maximum, minimum temperatures and rainfall 1980±1994.
tality association in Hong Kong. The objectives of this research are to test the hypothesis that there was seasonality of mortality for the period from 1980 through 1994 and to assess weather±mortality relationships for genders and various age groups.
Data and method Data Monthly mortality from various diseases in this study period were obtained from the Census and Statistics Department. The data were further divided into gender and age groups 0±16, 17±24, 25±44, 45±64 and r65. Current knowledge suggests that there are associations between various weather variables and mor-
tality. Weather variables including temperature, dewpoint, pressure, cloud, windspeed and precipitation were selected for this analysis and monthly meteorological data for the study period were obtained from the Hong Kong Observatory (Table 1). Figure 1 presents the average monthly maximum and minimum temperatures and rainfall for the study period. Temperatures were close to the normal. However, the summer rainfall was below normal. Rainfall anomalies were observed in the mean May and August precipitation ®gures. The May average of 356.4 mm exceeded the norm of 316.7 mm and the August values was 328.3 mm as compared with the norm of 391.4 mm. Determination of mortality seasonality There were a total of 410,281 deaths in the study area and the causes of death from neoplasm, circulatory and respiratory systems failure comprised 29.2, Table 2 Selected causes of death
Fig. 3. Monthly mortality of sex speci®c and total deaths from all causes.
Cause of death
ICD section
ICD codes
All causes Neoplasm (malignant) Circulatory system Hypertensive disease Ischaemic heart disease (IHD) Cerebrovascular disease (CVD) Other circulatory diseases Respiratory diseases
I±XVII II VII
inclusive 140±208 390±459 401±405 410±414 430±438
VIII
460±519
422
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
Table 3 Actual numbers of deaths of selected diseases of the circulatory system
Male Female Age 25±44 Age 45±64 Age r65
Hypertensive disease
IHD
CVD
Other circulatory
6672 6155 214 3319 9289
21,596 18,344 666 9077 30,163
22,394 24,674 1141 11674 33,979
9099 8981 1237 4733 11,936
28.8 and 16.8% respectively (Fig. 2). Only the major causes of death were selected for analysis (Table 2). The greatest numbers of deaths from circulatory system were the hypertensive disease, ischaemic heart disease (IHD) and cerebrovascular disease (CVD) (Table 3). Monthly gender and age speci®c and total mortality data were divided by their corresponding annual midyear population to eliminate the eect of population growth. Further, these adjusted monthly mortality data were then standardized to a 30.5 day average to correct for monthly dierences in number of days. Winter peaks existed in total and gender and age speci®c mortality from all causes (Figs. 3±5). In order to test the existence of mortality seasonality, dummy variables (winter months, i.e. January, February and December=1, while other months=0) were employed in regression analyses. ANOVA was then utilized to test for dierences in seasonal mortality. Determination of weather±mortality relationship Regression analysis was used to ascertain which combination of weather elements would produce the highest coecient determination (r 2) for age and gender speci®c deaths and total mortality from all causes and selected diseases. Each meteorological variable was
Fig. 4. Monthly mortality of sex speci®c deaths from selected causes.
Fig. 5. Monthly mortality of age r65 from selected causes.
checked by inspection of scatterplots for a nonlinear relation to mortality. When non-linearity was evident, appropriate transformation was employed. In this analysis, maximum temperature, minimum temperature and dewpoint were logarithmically transformed, whereas a square root transformation was chosen for precipitation. Then a stepwise approach was adopted. Regression diagnostics including residual plots and variance in¯ation factors (VIFs) (Draper and Smith, 1981) were used to choose the ®nal model. A high VIF indicates the inclusion of collinear independent variables in the model. When this occurs, one of the collinear variables is omitted. The ®nal selected model must have VIF for all independent variables less than 2.00 and randomly distributed residuals.
Results Mortality seasonality Results of the regression analyses using dummy variables (Table 4) revealed a positive relationship between mortality and the winter months, i.e. higher deaths occurred in winter. Statistically signi®cant reliable winter peaks in gender speci®c and total deaths from all causes, circulatory and respiratory diseases were noted. However, cancer mortality was not seasonal. There was no gender dierence in mortality seasonality. Excepting neoplasms, results from ANOVA (Table 5) exhibited signi®cant dierences of mortality among seasons and this further con®rmed the existence of a reliable winter peak for all causes and selected disease mortality. For the various age groups, no signi®cant seasonality for neoplasm was discovered. Only age groups 45± 64 and r65 displayed signi®cant winter peaks in deaths from circulatory and respiratory systems. Signi®cant seasonality for all causes of deaths existed only in age group r65. The winter and summer mortality rates for
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
423
Table 4 Models for mortality seasonality (winter peak) Total death
All causes Neoplasm Circulatory Hypertensive IHD CVD Other circulatory Respiratory
Male
Female
Age r65
regression coecient
p
regression coecient
p
regression coecient
p
regression coecient
p
6.34 0.26
0.00 0.14
6.44 0.33
0.00 0.17
6.23 0.19
0.00 0.57
73.71 3.03
0.00 0.07
0.50 1.06 1.04 0.51 1.76
0.00 0.00 0.00 0.00 0.00
0.46 1.09 1.04 0.48 2.05
0.00 0.00 0.00 0.00 0.00
0.55 1.05 1.01 0.56 1.46
0.00 0.00 0.00 0.00 0.00
5.11 11.56 10.78 6.22 18.76
0.00 0.00 0.00 0.00 0.00
the age speci®c deaths, presented in Table 6, further reveals the minimal seasonal mortality dierences among the younger age groups. Further, for the age groups below 44, the summer mortality rates from all causes were found to be slightly higher than the winter ones. The same fact happened in age group 25±44 for IHD, CVD and other circulatory causes. Weather±mortality relationship The results of the regression analyses were shown in Table 7. No signi®cant relationship was detected between weather variables and cancer mortality for total, gender and age speci®c deaths. Minimum temperature had the most signi®cant impact on deaths from the selected causes. For mortality from all causes, a signi®cant negative relationship between minimum temperature and a positive association between cloud cover and gender speci®c and total deaths were observed. However, the contributing weather variables aecting female mortality from all causes were slightly dierent from those of male mortality. Minimum temperature, cloud cover
and windspeed were included in the female model while the male model only contained minimum temperature. Minimum temperature and cloud had the most in¯uential eects on deaths from the circulatory system. Windspeed was discovered to have a signi®cant negative association with deaths from CVD. Deaths from respiratory diseases were also connected with cold and cloudy conditions. There was no prominent sex dierence in mortality from these causes. The evaluation of weather impact on mortality for various age groups revealed that no relationship was observed for the two youngest age groups (ranging from 0 to 24 yr old). For age group 25±44, deaths from IHD and other circulatory diseases were not weather related. The negative eect of windspeed was found to be more notable than the eect of minimum temperature as it was included in all models. The weather±mortality relationships were discovered to be weak (Adj r 2 values ranged from 0.05 to 0.07). Deaths for age groups 45±64 and r65 were mainly contributed by the combined force of minimum temperature and cloud cover. The negative in¯uence of windspeed was
Table 5 Results of ANOVA Total death
All causes Neoplasm Circulatory Hypertensive IHD CVD Other circulatory Respiratory
Male
Female
Age r65
F Ratio
p
F Ratio
p
F Ratio
p
F Ratio
p
63.86 0.91
0.00 0.35
47.87 1.19
0.00 0.32
66.55 0.67
0.00 0.57
45.81 1.88
0.00 0.13
31.16 40.26 49.23 31.17 46.78
0.00 0.00 0.00 0.00 0.00
22.65 32.14 41.05 26.15 37.51
0.00 0.00 0.00 0.00 0.00
31.31 34.15 39.64 27.58 48.12
0.00 0.00 0.00 0.00 0.00
31.57 62.24 33.61 23.76 55.63
0.00 0.00 0.00 0.00 0.00
424
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
Table 6 Mortality rates (per 100,000) for age speci®c groups All causes Age 0±16 Winter Summer Age 17±24 Winter Summer Age 25±44 Winter Summer Age 45±64 Winter Summer Age r65 Winter Summer
Hypertensive disease
IHD
CVD
Other circulatory
Respiratory
262.8 288.5
0.0 0.0
0.0 0.0
2.3 2.3
5.9 5.9
26.6 24.3
170.6 180.0
0.0 0.0
1.0 1.0
5.0 3.6
7.7 6.3
13.1 11.7
384.3 393.3
4.1 3.2
8.9 9.5
15.3 16.2
17.1 17.6
26.1 23.0
3059.6 2858.0
100.4 72.9
257.4 202.5
333.5 265.5
72.0 60.8
365.9 302.9
16356.6 12572.6
707.9 429.8
2108.7 1494.9
2338.7 1757.7
898.7 572.0
3641.4 2724.8
Fig. 6. Scatterplot of cerebrovascular (CVD) mortality (total death) versus minimum temperature.
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
425
Table 7 Models for selected causes of deatha Explanatory variables Total death MINT CLD WINDSP Adj r 2 Male MINT CLD WINDSP Adj r 2 Female MINT CLD WINDSP Adj r 2 Age 25±44 MINT CLD WINDSP Adj r 2 Age 45±64 MINT CLD WINDSP Adj r 2 Age r65 MINT CLD WINDSP Adj r 2 a b
All causes ÿ33.06 4.35 0.61 ÿ33.83 0.53 ÿ34.03 7.51 ÿ0.92b 0.62
Hypertensive disease ÿ2.64 0.36b 0.44 ÿ2.53 0.38b 0.35 ÿ2.83 0.49b 0.45
IHD ÿ5.36 0.48 ÿ5.49 0.41 ÿ5.27 0.44
CVD
Other circulatory
ÿ5.65 1.40 ÿ0.24 0.59
ÿ2.87 0.53
ÿ9.32 2.24
0.43
0.46
ÿ5.59 0.98b ÿ0.33 0.58
ÿ2.57 0.40b
ÿ5.61 1.84 ÿ0.30b 0.44
ÿ3.24 0.68b
0.33
0.41
ÿ0.12b ÿ0.70b 0.05
ÿ0.02b 0.05
ÿ27.07b
ÿ2.74 0.93b
0.04 ÿ378.14 45.04 0.49
0.17 ÿ28.00 4.84b ÿ1.14b 0.47
0.29 ÿ59.28 0.66
ÿ6.93 2.16b ÿ1.06 0.21 ÿ60.13 17.13 ÿ3.87b 0.48
ÿ10.85 2.47 0.44 ÿ8.09 2.67 0.44 ÿ0.34b 0.28b ÿ0.10b 0.06
ÿ0.08b 0.07 ÿ4.84
Respiratory
ÿ1.61b 0.92b
ÿ8.39 2.83b
0.06
0.24
ÿ32.32
ÿ101.47 24.24
ÿ1.77b 0.36
0.54
Numbers represent regression coecient (b) at p = 0.00. Numbers represent regression coecient (b) at p < 0.05.
found to be more prominent for age group r65. Much stronger weather±mortality associations were also detected in the elderly (Adj r 2 values ranged from 0.36 to 0.66).
Discussion The ®ndings that mortality seasonality existed for gender speci®c and total deaths are consistent with previous studies (Bako et al., 1988; Douglas et al., 1991a, 1991b). Although Allan (1966) found that the highest mortality from cancer in Britain was not in the coldest months, but in the fourth quarter of the year, the result of this study indicated no seasonality for malignant disease. This ®nding con®rms the absence of a winter peak. The non-existence of signi®cant seasonality for the
younger age groups (0±44) was further supported by the ®ndings of this study that deaths of those groups (0±24) were not weather related. Further, this also evidences that the younger age groups are less vulnerable to weather eects. For the weather±mortality relationships, it is not surprised that maximum temperature was excluded from all models. Majority of the population have acclimatized the hot, sultry summer in the subtropics. Any temperature changes, particularly cold weather change, will have more obvious eect and trigger more deaths (Makino, 1993; Auliciems et al., 1997). The negative relationship and positive connection between minimum temperature and cloud cover respectively and mortality from all causes suggest that overcast and colder days may provide conditions for increased mortality. Windspeed was not included in all the models. This shows that the stressful eect of wind
426
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427
in colder weather has less in¯uence on mortality as expected and con®rms the previous ®ndings by Kalkstein and Davis (1989). It is not unexpected to discover no prominent gender dierence in mortality from the various causes. However, the inclusion of windspeed and cloud cover in the female model from all causes of death indicated that females were more sensitive to the combined eect of weather stress. Minimum temperature and cloud were the contributing factors to deaths from circulatory system failures. The ®nding of a negative association between windspeed and CVD mortality further reveals the eect of wind on mortality is negligible. Minimum temperature was the most in¯uential factor. It is found that exposure to the cold increases blood pressure, blood viscosity and heart rate (Kunst et al., 1993) and may precipitate circulatory diseases. Pan et al. (1995) stated that there was a U-shaped relation between temperature and cardiovascular mortality in Taiwan. However, no such association was observed in Hong Kong (Fig. 6). Deaths from respiratory diseases were connected with minimum temperature and cloud cover. The eect of colder temperature on the respiratory system has an important role. The alveoli are embedded in a capillary network. The alveolar cells and the capillary endothelial cells comprise two layers of cells between the air and blood. A continuous sheet of blood is close to the alveolar wall. The lungs are exposed directly to the environment without the protection of skin. This intimate environmental contact induces not only respiratory diseases but also provokes changes in circulatory and other diseases (Douglas et al., 1991b). The strong weather±mortality associations in age group r65 and no or very weak relationships for the three youngest age groups (ranging from 0 to 44 yr old) are important discoveries. They are consistent with previous research ®ndings that the elderly are more susceptible to weather stress. This apparent weather±mortality relationship is due to the failure of homeostatic defence mechanism with advancing age that facilitates the onset of hyperthermia or hypothermia, which in turn would trigger circulatory and other diseases (Douglas et al., 1991b; Pan et al., 1995). Further, Woo et al. (1991) also found that the elderly might be more vulnerable to less extreme temperature changes in subtropical climate. Conclusions It is apparent that weather has a substantial eect on mortality. The impact can dier seasonally. Also certain age groups seem to be more susceptible to the in¯uence of weather. Results of this study con®rm some previous ®ndings, but also contradict other ®nd-
ings in weather±mortality research. More attention has recently been placed on the impact of air pollution on mortality and many studies suggest that pollution has a signi®cant eect on health (Stern, 1977). More promising results would be expected if air pollution variables are included. The results also indicate that the situation in Hong Kong is not the same as other tropical or subtropical regions. Rainfall was not associated with mortality as in Costa Rica or Sri Lanka (Madrigal, 1994; Motohashi et al., 1996). Humidity was found to be unrelated to mortality unlike Brisbane (Auliciems et al., 1997). Further research in these regions should be conducted to clarify the complexity of weather±mortality relationships. These research ®ndings contribute to the formulation of public health policy, such as justifying increasing resources for elderly care. Knowledge of weather±mortality relationships may provide better insights on which diseases are more likely to cause death under certain weather conditions and the susceptibility of speci®c age groups. This type of research can aid in the development of health-related weather services to the general public, such as indices to warn of severe temperature or pollution episodes. References Alberdi, J.C., Diaz, J., Montero, J.C., Miron, I., 1998. Daily Mortality in madrid community 1986±1992: relationship with meteorological variables. European Journal of Epidemiology 14, 571±578. Allan, T.M., 1966. Seasonal Distribution of Deaths from Cancer. British Medical Journal 10, 837±841. Applegate, W.B., Runyan Jr., J.W., Bras®eld, L., Willians, M.L., Konigsberg, C., Fouche, C., 1981. Analysis of 1980 the heat wave in Memphis. Journal of American Geriatrics Society 29, 337±342. Auliciems, A., Frost, D., Siskind, V., 1997. The time factor in mortality: weather association in a subtropical environment. International Journal of Biometeorology 40, 183± 191. Bako, G., Ferenczi, L., Hill, G.B., Lindsay, J., 1988. Seasonality of mortality from various diseases in Canada 1979±83. Canadian Journal of Public Health 7, 388±389. Bridger, C.A., Ellis, F.P., Taylor, H.L., 1976. Mortality in St Louis during heat waves in 1936, 1953, 1954, 1955 and 1966. Environmental Research 12, 38±48. Buechley, R.W., Van Bruggen, J., Truppi, L.E., 1972. Heat Island=Death Island? Environmental Research 5, 85±92. Christophersen, O., 1997. Mortality during the 1996/7 winter. Population Trend 90, 11±17. Donaldson, G.C., Keatinge, W.R., 1997. Mortality related to cold weather in elderly people in Southeast England 1979± 94. British Medical Journal 315, 1055±1056. Douglas, A.S., Al-Sayer, H., Rawles, J.M., Allan, T.M., 1991a. Seasonality of disease in Kuwait. Lancet 337, 1393± 1397.
Y.Y. Yan / Social Science & Medicine 50 (2000) 419±427 Douglas, A.S., Allan, T.M., Rawles, J.M., 1991b. Composition of seasonality of diseases. Scottish Medical Journal 36, 76±82. Draper, N., Smith, H., 1981. Applied Regression Analysis. John Wiley, New York. Ellis, F.P., 1972. Mortality from heat illness and heat-aggravated illness in the US. Environmental Research 5, 1±58. Ellis, F.P., Nelson, F., 1978. Mortality in the elderly in a heat wave in New York City, August 1975. Environmental Research 15, 504±512. Gerald, F., Rose, R., 1979. Blizzard morbidity and mortality: Rhode Island, 1978. American Journal of Public Health 69, 1050±1052. Gover, M., 1938. Mortality during periods of excessive Temperature. US Public Health Report 53, 1122. Hodge, W., 1978. Weather and mortality. In: EDIS. US Department of Commerce, Washington. Kalkstein, L.S., Davis, R.E., 1989. Weather and human mortality: an evaluation of demographic and interregional responses in the US. Annals of the Association of American Geographers 79, 44±64. Kalkstein, L.S., Smoyer, K.E., 1993. The impact of climate change: some international implications. Experientia 49, 969±979. Knobeloch, L., Anderson, H., Morgan, J., Nashold, R., 1997. Heat-related illness and death, Wisconsin, 1995. Wisconsin Medical Journal 96, 33±38. Kunst, A.E., Looman, C.W.N., MacKenbach, J.P., 1993. Outdoor air temperature and mortality in the netherlands: a time-series analysis. American Journal of Epidemiology 137, 331±341. Larson, U., 1990. Short-term ¯uctuation in death by cause, temperature and income in the United States, 1930±1985. Social Biology 37, 172±187. Lerchl, A., 1998. Changes in the seasonality of mortality in Germany from 1946 to 1995: the role of temperature. International Journal of Biometeorology 42, 84±88. Madrigal, L., 1994. Mortality seasonality in EscazuÂ, Costa Rica 1851±1921. Human Biology 66, 433±452. Makino, K., 1993. Weather/season and death. Asian Medical Journal 36, 580±587.
427
Marmor, M., 1975. Heat wave mortality in New York City, 1949 to 1970. Archive of Environmental Health 30, 130±136. Mather, J.R., 1974. Climatology: Fundamental and Application. McGraw-Hill, New York. McKee, M., Sanderson, C., Chenet, L., Vassin, S., Shkolnikov, V., 1998. Seasonal variation in mortality in Moscow. Journal of Public Health and Medicine 20, 268± 274. Motohashi, Y., Takano, T., Nakamura, K., Nakate, K., Tanaka, M., 1996. Seasonality of Mortality in Sri Lanka: biometeorological considerations. International Journal of Biometeorology 39, 121±126. Oliver, J.E., 1981. Climatology: Selected Applications. John Wiley, New York. Pan, W.H., Li, L.A., Tsai, M.J., 1995. Temperature extremes and mortality from coronary heart disease and cerebral infarction in elderly Chinese. Lancet 345, 353±355. Ramlow, J.M., Kuller, L.H., 1990. Eects of the summer heat wave of 1988 on daily mortality in Allegheny County, PA. Public Health Reports 105, 283±289. Saez, M., Sunyer, J., Castellsague, J., Murillo, C., AntoÂ, J.M., 1995. Relationship between weather temperature and mortality: a time series analysis approach in Barcelona. International Journal of Epidemiology 24, 576±582. Shumway, R.H., Azari, A.S., Pawitan, Y., 1988. Modeling mortality ¯uctuation in los angeles as function of pollution and weather eects. Environmental Research 45, 224±241. Smoyer, K.E., 1998. A comparative analysis of heat waves and associated mortality in St. Louis, Missouri 1980 and 1995. International of Biometeorology 42, 44±50. Stern, A.C., 1977. The Eects of Air Pollution. Academic Press, New York. Tromp, S.W., 1963. Medical Biometeorology. Elsevier, New York. Whitman, S., Good, G., Donoghue, E.R., Benbow, N., Shou, W., Mou, S., 1997. Mortality in Chicago attributed to the July 1995 heat wave. American Journal of Public Health 87, 1515±1518. Woo, J., Kay, R., Nicholls, M.G., 1991. Environmental temperature and stroke in a subtropical climate. Neuroepidemiology 10, 260±265.